Departamento de Estadística
http://hdl.handle.net/10016/12
2016-05-28T04:02:59ZSmall area estimation of general parameters under complex sampling designs
http://hdl.handle.net/10016/22731
Small area estimation of general parameters under complex sampling designs
Guadarrama, María; Molina, Isabel; Rao, J.N.K.
Universidad Carlos III de Madrid. Departamento de Estadística
When the probabilities of selecting the individuals for the sample depend on the
outcome values, we say that the selection mechanism is informative. Under informative
selection, individuals with certain outcome values appear more often in the sample and
therefore the sample is not representative of the population. As a consequence, usual
model-based inference based on the actual sample without appropriate weighting might
be strongly biased. For estimation of general non-linear parameters in small areas, we
propose a model-based pseudo empirical best (PEB) method that incorporates the
sampling weights and reduces considerably the bias of the unweighted empirical best
(EB) estimators under informative selection mechanisms. We analyze the properties of
this new method in simulation experiments carried out under complex sampling
designs, including informative selection. Our results confirm that the proposed weighted
PEB estimators perform significantly better than the unweighted EB estimators in terms
of bias under informative sampling, and compare favorably under non-informative
sampling. In an application to poverty mapping in Spain, we compare the proposed
weighted PEB estimators with the unweighted EB analogues.
2016-04-01T00:00:00ZRemittances in Mexico and their unobserved components
http://hdl.handle.net/10016/22674
Remittances in Mexico and their unobserved components
Corona, Francisco; Orraca, Pedro
Universidad Carlos III de Madrid. Departamento de Estadística
The present study aims to determine the common trends and the permanent and
transitory components of remittances received by Mexican households. This is
done by estimating a small Dynamic Factor Model (DFM), using the approach
first proposed by Gonzalo and Granger (1995), determining the number of
common trends subject to the cointegration results. The study also shows the
similarities between this small DFM with respect to large DFM, which are
widely used in the econometric literature. The results indicate the presence of
one cointegration relationship. Consequently, there are four common trends.
These common factors are negatively dominated by Mexico's economic
activity and positively by the U.S. industrial production. The effects of the
exchange rate and the U.S. unemployment rate are positive, but less relevant.
This economic scenario leads to remittances exceeding its permanent
component
2016-04-01T00:00:00ZUsing high-frequency data and time series models to Improve yield management
http://hdl.handle.net/10016/3114
Using high-frequency data and time series models to Improve yield management
Cancelo, José Ramón; Espasa, Antoni
High-frequency (less than monthly) time series data provide valuable information for designing the adequate yield policy of the organisation. However, it is not easy to extract this information from raw data; although the evolution of the series is usually induced by stable patterns of behaviour of the economic agents, these patterns are so complex that simple smoothing techniques or subjective forecasting cannot consider all underlying factors. In this paper, we discuss time series models as a tool for carrying out a full and efficient analysis. The main ideas are illustrated with an application to Spanish daily electricity consumption.
2001-01-01T00:00:00ZA Partial parametric path algorithm for multiclass classification
http://hdl.handle.net/10016/22390
A Partial parametric path algorithm for multiclass classification
Liu, Ling; Martín Barragán, Belén; Prieto, Francisco J.
Universidad Carlos III de Madrid. Departamento de Estadística
The objective functions of Support Vector Machine methods (SVMs) often includeparameters to weigh the relative importance of margins and training accuracies.The values of these parameters have a direct effect both on the optimal accuraciesand the misclassification costs. Usually, a grid search is used to find appropriatevalues for them. This method requires the repeated solution of quadraticprograms for different parameter values, and it may imply a large computationalcost, especially in a setting of multiclass SVMs and large training datasets. Formulti-class classification problems, in the presence of different misclassificationcosts, identifying a desirable set of values for these parameters becomes evenmore relevant. In this paper, we propose a partial parametric path algorithm, basedon the property that the path of optimal solutions of the SVMs with respect tothe preceding parameters is piecewise linear. This partial parametric path algorithmrequires the solution of just one quadratic programming problem, and anumber of linear systems of equations. Thus it can significantly reduce the computationalrequirements of the algorithm. To systematically explore the differentweights to assign to the misclassification costs, we combine the partial parametricpath algorithm with a variable neighborhood search method. Our numerical experimentsshow the efficiency and reliability of the proposed partial parametricpath algorithm.
2016-02-01T00:00:00Z